What is it about?
In this paper, we present a novel framework that leverages DNNs to search for "counterexamples" - scenarios where a CPS might malfunction and cause harm. This approach utilizes the power of existing adversarial attack algorithms designed for DNNs to uncover hidden safety risks.
Featured Image
Photo by Kai Pilger on Unsplash
Why is it important?
Connecting DNNs and CPS: We establish a link between falsification of safety properties in CPS and falsification of DNNs. New Framework: Our framework utilizes a DNN model of the CPS and applies DNN falsification tools to find vulnerabilities. Effective for Various Systems: While applicable to general systems, we showcase its effectiveness on CPS with both linear and non-linear dynamics. Identifying Hard-to-Find Issues: We demonstrate the ability to detect previously unknown safety risks in CPS.
Perspectives
This research can help improve the safety and reliability of CPS in various critical domains like healthcare, transportation, and automation.
Sauvik Gon
Read the Original
This page is a summary of: Data-Driven Falsification of Cyber-Physical Systems, February 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3641399.3641401.
You can read the full text:
Contributors
The following have contributed to this page
